This study investigates the morphometric variation of osteoderms across species and populations within the genus Acontias, especially in the A. meleagris species complex. Using both univariate and multivariate analyses, we evaluated whether size-independent osteoderm morphometrics could effectively differentiate taxa, particularly where minimal genetic divergence is present (i.e., cases where morphologically diagnosable species show little genetic separation). Univariate analysis revealed some significant morphometric variation across osteoderm regions, with sub-regions C and D showing the highest discriminatory power. Multivariate analyses, including principal component analysis (PCA) and discriminant function analysis (DFA), demonstrated the complementary strengths of ratio-based and studentized linearized residuals (SLRs)-based metrics. Ratio-based analyses were more effective in distinguishing genetic species, while SLRs-based analyses captured finer-scale population differences. The combined approach improved classification accuracy, underscoring the value of integrating multiple morphometric methods. Our results suggest that osteoderm morphometrics provide valuable supplementary data for species delimitation and may help resolve taxonomic boundaries within Acontias and possibly other lizards. However, the limited ability to differentiate morphs and populations in some cases highlights the need for additional data, such as environmental or behavioral traits. The findings have the potential to improve taxonomic resolution among skinks and contribute to broader taxonomic frameworks in herpetological systematics.

Legless lizard, morphology, multivariate analysis, size-independent, skink, species delimitation, studentized linearized residuals, ratios, reptile

Osteoderms are plates comprising mainly calcium phosphate and collagen found in the epidermal scales of many tetrapods, including crocodilians, squamates, certain frogs, armadillos, and some extinct taxa (Boulenger 1887; Camp 1923; Oliver 1951; Strahm and Schwartz 1977; Hill 2005; Williams et al. 2022). These structures exhibit significant variation in number, shape, and sculpturing and may therefore potentially be of use as diagnostic characters in taxonomic studies. Closer examination of their utility has, however, led to varying views. For instance, Boulenger (1887) argued that osteoderms were too variable at the intraspecific level to be taxonomically informative, while others, including Camp (1923), Hewitt (1929), Hoffstetter (1962), and Lambiris (1992), considered them valuable taxonomic markers. Camp (1923) notably emphasized the importance of comparing osteoderms from identical body regions to ensure diagnostic accuracy. Strahm and Schwartz (1977) found them to be diagnostic up to genus level in the Diploglossinae and Anguidae, while Burns (2008) demonstrated their taxonomic relevance in ankylosaurid and notosaurid ornithischian fossil taxa. Osteoderms have been attributed various functions beyond their taxonomic significance, including defense against predators and protection during intraspecific combat (Vickaryous et al. 2015). Furthermore, Broeckhoven et al. (2015) experimentally demonstrated that osteoderms provide protection by reducing skin penetration from simulated bites of different mongoose species, supporting their role in predator defense. Additional proposed roles include thermoregulation, nutrient storage, and water retention (Williams et al. 2022). The thermoregulatory function of osteoderms has been linked to their vascularization, although this remains debated, with most studies focusing on crocodilians and turtles rather than squamates (Williams et al. 2022).

In lizards, comparative lepidosis remains the primary method for taxonomic diagnosis and is expected to continue being an indispensable component of taxonomic keys due to its practicality and wide application. Prominent taxonomic references for skinks, such as those by FitzSimons (1943), Broadley and Greer (1969), and Branch (1998), are all primarily based on lepidosis. However, the diagnostic potential of osteoderm polymorphism, particularly in skinks, remains underexplored. Skink osteoderms are relatively simple in structure and morphology (see fig. 3 in Williams et al. (2022) for detailed descriptions of Scincidae), making them easier to quantify and analyze compared to the more complex osteoderms found in anguid lizards. For detailed structural descriptions of Anguidae, see figs 1, 2 in Strahm and Schwartz (1977). This simplicity offers an advantage for morphometric comparisons, facilitating the testing of osteoderms as a reliable diagnostic feature among skinks. Such testing could yield significant insights, not only within skinks but also across other lizard taxa that exhibit similarly straightforward osteoderm structure. The taxonomic importance of osteoderms lies in their potential to enhance phylogenetic analyses by integrating multiple diagnostic traits. Recent findings by Daza et al. (2024) have demonstrated that osteoderms are not only morphologically informative but also provide evolutionary insights. These findings highlight the role of osteoderms in elucidating phylogenetic relationships among both extant and extinct skink species. As a result, osteoderms could serve as a crucial trait in systematics, complementing other morphological and molecular data, and thereby strengthening the resolution of phylogenetic hypotheses across lizard taxa.

The taxonomy and species boundaries within the morphologically conservative legless skink subfamily Acontinae have long been considered ambiguous and challenging to delineate (Zhao et al. 2023). Molecular evidence has recently shed light on these complexities, revealing that Acontias occidentalis is closely related to Acontias percivali, with the latter being nested within the A. occidentalis species complex (Zhao et al. 2023). Furthermore, despite Acontias lineicauda showing a close phylogenetic relationship to the Acontias meleagris species complex (Daniels et al. 2005, 2009; Engelbrecht et al. 2013), the genetic distance between A. lineicauda and other members of the A. meleagris complex remains significant (Zhao et al. 2023). Further molecular analyses have resulted in the synonymization of previously recognized species, such as Acontias percivali tasmani and Acontias orientalis, with A. meleagris (Daniels et al. 2005, 2009; Engelbrecht et al. 2013; Zhao et al. 2023). Moreover, considerable genetic divergence has been detected among lineages within the A. meleagris complex. For instance, individuals from the Coega area, which were identified morphologically as “A. p. tasmani”, and geographically distant populations from Mossel Bay were found to belong to the same clade. In contrast, populations of A. meleagris from Robben Island and Saldanha Bay form a distinct, separate clade (Daniels et al. 2005, 2009; Engelbrecht et al. 2013). These findings emphasize the intricate evolutionary history and taxonomic complexity within the Acontinae, underscoring the need for integrative approaches combining molecular and morphological data.

In morphometric taxonomy, both studentized linearized residuals (SLRs) and ratios are integral to multivariate analyses such as principal component analysis (PCA) and discriminant function analysis (DFA) (Rohlf 1990; Zhao et al. 2021). Each approach offers distinct advantages and limitations, making them appropriate for different research contexts. Ratios are especially useful for capturing relative differences between measurements, which is crucial for differentiating species or populations based on body proportions. They are widely employed in morphometric studies, particularly in taxonomy, where body size comparisons are pivotal (Baur and Leuenberger 2011). Ratios facilitate the scaling of body size effects and can be interpreted within the framework of allometry, which pertains to shape changes related to size. However, a significant drawback of ratios is their sensitivity to size differences, which can sometimes obscure true shape variation by conflating size-related and shape-related differences. Additionally, when ratios are directly applied in multivariate methods like PCA, they may not fully capture complex shape relationships unless appropriately adjusted for size (Courtenay 2023). SLRs, on the other hand, are derived by adjusting the residuals from regression analyses for variability, making them a robust tool for size correction. They are particularly valuable in controlling for body size effects, providing a clearer representation of shape variation independent of size. SLRs are advantageous when the focus is on shape differences without the confounding influence of size (Baur and Leuenberger 2011; Lehmann 2012). Nonetheless, SLRs also present challenges, including their computational and interpretative complexity compared to ratios, which requires a more profound statistical understanding. Moreover, in datasets with high variability, even well-standardized residuals may not entirely negate the influence of size (Baur and Leuenberger 2011; Courtenay 2023).

In this study, we aimed to evaluate the taxonomic utility of osteoderms in distinguishing among several fossorial skink species: Acontias occidentalis, A. percivali, A. lineicauda, A. meleagris, and the morphs “A. orientalis” and “A. p. tasmani” (Lamb et al. 2010). The distribution of these species is as follows: A. meleagris is found along the southern coast of South Africa; A. occidentalis occurs in parts of Angola, Namibia, Botswana, Zimbabwe, and South Africa; A. percivali is restricted to southeastern Kenya; A. lineicauda is native to the area around Port Elizabeth and Port Alfred in the Eastern Cape Province of South Africa; and the “A. p. tasmani” and “A. orientalis” morphs are found in the Eastern Cape Province (Broadley and Greer 1969). We hypothesized that (i) distinct species could be reliably differentiated based on osteoderm morphometrics from corresponding body regions, with closely related morphospecies exhibiting smaller morphometric distances compared to more distantly related ones; (ii) osteoderm characteristics would be invariant (fixed) within populations of the A. meleagris species complex; and (iii) both ratios and SLRs would provide informative morphometric data in multivariate analyses to differentiate among taxa.

Specimens from the following localities were analyzed, with a consistent sample size of ten per group: Acontias occidentalis (Windhoek, Namibia [22.5609°S, 17.0658°E]), Acontias percivali (Voi, Kenya [3.3961°S, 38.5633°E]), and “Acontias p. tasmani” (Coega, Eastern Cape Province, South Africa [33.8020°S, 25.6883°E]). Additionally, Acontias lineicauda (Port Elizabeth, Eastern Cape Province, South Africa [33.7174°S, 25.8142°E]) and “Acontias orientalis” (Port Elizabeth, Eastern Cape Province, South Africa [33.7521°S, 25.6995°E]) were included (Fig. 1). Furthermore, samples were obtained from three geographically distinct populations of Acontias meleagris in the Western Cape Province (Fig. 1), with the following sample sizes: Mossel Bay ([34.1831°S, 22.1460°E], n = 10), Robben Island ([33.8066°S, 18.3662°E], n = 5), and Saldanha Bay ([33.0117°S, 17.9443°E], n = 8). A GIS map was generated using QGIS Geographic Information System (QGIS Development Team, 2023), version 3.30.2, employing the Natural Earth II 10 m raster dataset (available at: https://www.naturalearthdata.com/downloads/10m-raster-data/10m-natural-earth-2/). All spatial data were projected using the WGS 1984 coordinate reference system. All specimens had a snout-vent length greater than 80% of the maximum published for the species (Branch 1998) and were therefore considered adults.

Figure 1. 

Map illustrating the sampling locations for this study. Abbreviations: “AMO” = Acontias orientalis, “AMMM” = Acontias meleagris (Mossel Bay population), “AML” = Acontias lineicauda, “AMMS” = A. meleagris (Saldanha Bay population), “AMMR” = A. meleagris (Robben Island population), “APO” = Acontias occidentalis, “APP” = Acontias percivali, and “APT” = Acontias percivali tasmani. The map was generated using QGIS v3.30.2 (QGIS Development Team) with the Natural Earth II 10 m raster dataset (available at Natural Earth: https://www.naturalearthdata.com/downloads/10m-raster-data/10m-natural-earth-2/). Blue lines represent rivers, while black lines indicate country borders.

Scales were collected from five body regions: the mid-dorsal body (MDB), mid-ventral body (MVB), the pre-anal scale (PA), mid-dorsal tail (MDT), and mid-ventral tail (MVT). The complete set of osteoderms was extracted from each specimen. For species with a sample size of 10, this meant that one MDB osteoderm was removed from each individual, resulting in a total of 10 MDB osteoderms. The same approach was applied to all osteoderm types, ensuring a consistent sample size of 10 for each category. The preparation of scales followed the method developed by Lambiris (1992). This process involves first severing Sharpey’s fibers, which anchor the osteoderm between adjacent ones, by carefully sliding a fine-pointed blade across its anterior dorsal and anterior ventral sides. The osteoderm is then grasped posteriorly with fine forceps and gently pulled in a caudal direction for removal. Extracted osteoderms were stored individually in tubes containing alcohol. Scales were cleaned by macerating in 5% KOH and carefully removing connective tissue with fine forceps. They were then stained with a solution of 0.5M Alizarin Red in 5% KOH until the osteoderms were sufficiently highlighted. After staining, scales were dehydrated through sequential immersions in 70%, 96%, and 100% ethanol (EtOH), each for three minutes, followed by five minutes in xylene. The scales were mounted on microscope slides but were not cover-slipped. A diagram illustrating the basic structure of an acontine scale’s osteoderms is presented in Fig. 2. The osteoderms of each scale were subdivided into four sub-regions (sub-regions A–D) as shown in Fig. 3. In the case of the PA scale, this was, however, not possible as its osteoderms could not be readily grouped. Size and area determinations were made by scanning the scales to a computer and analyzing them using the Soft Imaging System analySIS Program (Soft Imaging System GmbH Master). For the PA scale, only the total area was recorded (for the measurement data, see Suppl. material 1).

Figure 2. 

Illustration of the osteoderm anatomical structure of the acontine skink (represented by Acontias orientalis), with scale units indicated.

Figure 3. 

Illustration of the four sub-regions (A–D) into which each osteoderm (represented by Acontias orientalis) was divided, with scale units indicated.

We normalized the dataset using a log transformation in R version 4.2.2 (R Core Team 2022). To assess and confirm the normality of the dataset, we applied the Kolmogorov-Smirnov test using the R package ‘nortest’ (Gross and Ligges 2015), and all log-transformed variables yielded a p-value greater than 0.05 (no significant deviations from the normal distribution), indicating significant improvement in data normality. For the statistical analyses, we first identified informative variables for distinguishing among groups through one-way analysis of variance (ANOVA) as a univariate approach, using the R packages ‘car’ (Fox 2023) and ‘DescTools’ (Signorell et al. 2019) in R version 4.2.2. To adjust for multiple comparisons, we applied the Bonferroni correction post-hoc test, which rebalances the compounding risk of Type I error by adjusting p-values. Two independent ANOVA analyses were conducted: (1) ANOVA based on the log-transformation of each variable (without accounting for body size variation), and (2) ANOVA based on ratios, where each variable representing areas of different sub-regions of the four anatomical regions (MDB, MVB, MDT, and MVT) was divided by the area of the PA (to account for proportional size and the effect of body size variability). We also performed analysis of covariance (ANCOVA) to assess the effect of body size using PA as the covariate, applying the R packages ‘emmeans’ (Lenth 2023) and ‘car’ (Fox 2023). The Bonferroni correction was again applied for post-hoc pairwise comparisons to evaluate significant differences between groups.

To test the first and second hypotheses, we employed multivariate analyses—PCA and DFA—to evaluate whether osteoderm area morphometrics could effectively differentiate between groups. First, PCA was used to reduce dimensionality through the R packages ‘factoextra’ (Kassambara 2016) and ‘factoMineR’ (Lê et al. 2008), allowing us to explore whether the reduced dimensions could reveal any diagnostic separation of taxa or populations without bias from predefined groupings (Courtenay 2023). We then used DFA, which incorporates grouping bias (Baur and Leuenberger 2011; Courtenay 2023), to assess whether each taxon or population could be correctly classified based on osteoderm morphometric data, utilizing the R packages ‘caret’ (Kuhn 2015) and ‘mass’ (Venables and Ripley 2002). To test the third hypothesis, we conducted PCA and DFA independently on (1) studentized linearized residuals (SLRs) of all variables (after transformation in R version 4.2.2) and (2) ratios of all variables divided by PA. Both approaches were employed to account for body size effects and to either remove or reduce these effects (Baur and Leuenberger 2011; Courtenay 2023). Lastly, we performed a combined PCA and DFA analysis using an integrative dataset that included both ratios and SLRs. During the DFA analysis, we applied the “remove-collinear-vars” function to eliminate highly collinear variables, using a threshold of 0.95 as the selection criterion.

This study examined the morphometric variation of osteoderms across four subregions of four body regions in selected species and populations of the genus Acontias, with a focus on distinguishing genetic species and populations within the A. meleagris species complex. Our primary objective was to determine whether osteoderm morphometrics could effectively differentiate between taxa and populations, consistent with genetic divergence patterns. We hypothesized that size-independent osteoderm metrics would offer significant discriminatory power among taxa and populations, even in cases of limited genetic divergence.

This study highlights the significant potential of osteoderm morphometrics for species delimitation and population differentiation within the legless lizard genus Acontias. Our findings underscore the importance of integrating multiple morphometric approaches to capture both broad-scale and fine-scale variation, particularly in taxa where genetic divergence may be minimal. By demonstrating the complementary strengths of ratio-based and SLRs-based methods, we provide a robust framework for future studies on reptilian morphometrics and evolutionary biology. Ultimately, our results contribute to a growing body of literature advocating for the use of morphometric data alongside genetic and ecological information to resolve taxonomic relationships in morphologically variable taxa. The findings have the potential to improve taxonomic resolution among skinks and contribute to broader taxonomic frameworks in herpetology.

We gratefully acknowledge the University of the Free State (UFS) for providing research facilities and funding through grant IFR2011041300046, which supported this project. We also extend our thanks to the South African Department of Environment and Nature Conservation (Permit No. FAUNA 624/2015) and the Department of Economic Development, Environment, and Tourism (Permit No. 001-CPM402-00001) for issuing the necessary collecting permits. We appreciate the ethical clearance provided by the UFS Ethical Committee (Animal Experiment No. NR04/2012 and NR07/2015), which was essential for the conduct of this research.